The proposed research examines regional dialect variation and sound change over time. The specific aims are to propose one explanation of why vowels change over generations and to advance our knowledge of variation in speech. A hypothesis is tested that vowel changes result, in part, from prosodic organization of speech. Accordingly, prosodic prominence will affect in predictable ways different vowel shifts currently underway in three regional varieties of American English found in Central Ohio, Southern Wisconsin, and Western North Carolina. The subject pool will consist of three generations of speakers who have lived in the same general area most of their lives. These will be children aged 8-11 years, their parents/caretakers aged 25-40 years, and their grandparents (or other members of this older generation) aged 55-75 years. Both production and perception experiments will be conducted that examine how prosodic prominence influences the acoustics of selected vowels and the extent to which listeners in a particular dialect (and from a particular generation) are sensitive to the acoustic changes induced by those systematic prosodic variations. The selected acoustic variables will measure the relationship between the formant pattern, magnitude and rate of formant frequency change and the degree of prosodic prominence (high, intermediate, and low). According to our hypothesis, the role of high prosodic prominence will be that of leading the vowel shift in a predictable direction. Detailed knowledge about the nature of normal and systematic variation in a given dialect is necessary for the development of clinical instruments for the fair and accurate assessment of both speech language and hearing disorders in various regions (and cultures) of the United States. The studies proposed here will contribute directly to development of these instruments. As regional varieties of English continue to diverge, this dimension must be incorporated into work on speech intelligibility, recognition, and synthesis. [unreadable] [unreadable]